论文标题
Web视图:用于描述Web浏览性能和交付的测量平台
Web View: A Measurement Platform for Depicting Web Browsing Performance and Delivery
论文作者
论文摘要
Web浏览是主要的Internet服务,每个客户都希望使用最少的网页加载时间。为了满足这种永久需求,网络浏览器提供新的处理引擎和增加功能。 Web开发人员利用了新的编程语言和范式,引入了新的传输协议,网络运营商提供了增加的带宽和内容输送网络(CDN)提供商部署最接近最终用户的Web服务器。为了评估这些发展的效率,通常将交付的性能评估为独立的方式,而无需考虑整个Web浏览链。我们提出了Web View,一个测量平台,该平台在用户代表性的环境中在流行网站上执行自动化的Web浏览会话。通过不同的可配置参数,Web View测量了一系列信息,以评估不同参数(传输协议,Web浏览器,CDN,访问网络等)的影响。基于这些测量值,为了简化对Web浏览的理解和相应的演变,Web视图提供了公共可视化网站(https://webview.orange.com)。例如,使用Web视图,我们能够检测到一天中不同时间使用不同CDN的使用可以将加载时间降低到400%,并且通过不同的传输协议交付内容可以将加载时间降低到79%。本文介绍了网络视图测量平台以及我们在一年以上测量中注意到的非凡事件。
Web browsing is the main Internet Service and every customer wants the minimum web page loading times. To satisfy this perpetual need, web browsers offer new processing engines and increased functionalities. Web developers make use of new programming languages and paradigms, new transport protocols have been introduced, network operators offer increased bandwidth and Content Delivery Networks (CDN) providers deploy web servers closest to end-users. To assess the efficiency of these evolutions, delivered performance is often evaluated into a standalone manner, without considering the whole chain of web browsing. We propose Web View, a measurement platform, which performs automated web browsing sessions on popular websites in a user-representative environment. Through different configurable parameters, Web View measures a wide set of information in order to evaluate the impact of the different parameters (transport protocols, web browsers, CDN, access networks, etc.). Based on those measurements, to ease the understanding of web browsing and the corresponding evolutions, Web View offers a public visualization website (https://webview.orange.com). For instance, with Web View, we were able to detect that the use of different CDNs at different times of the day can decrease loading times up to 400% and that the delivery of content through different transport protocols can decrease loading times up to 79%. This paper presents the Web View measurement platform as well as remarkable events we noticed during more than one year of measurements.